Medical Applications of Finite Mixture Models

Patients are not alike! This simple truth is often ignored in the analysis of me- cal data, since most of the time results are presented for the “average” patient. As a result, potential variability between patients is ignored when presenting, e.g., the results of a multiple linear regression model. In medicine there are more and more attempts to individualize therapy; thus, from the author’s point of view biostatis- cians should support these efforts. Therefore, one of the tasks of the statistician is to identify heterogeneity of patients and, if possible, to explain part of it with known explanatory covariates. Finite mixture models may be used to aid this purpose. This book tries to show that there are a large range of applications. They include the analysis of gene - pression data, pharmacokinetics, toxicology, and the determinants of beta-carotene plasma levels. Other examples include disease clustering, data from psychophysi- ogy, and meta-analysis of published studies. The book is intended as a resource for those interested in applying these methods.

The book shows how to model heterogeneity in medical research with covariate adjusted finite mixture models. The areas of application include epidemiology, gene expression data, disease mapping, meta-analysis, neurophysiology and pharmacology.

After an informal introduction the book provides and summarizes the mathematical background necessary to understand the algorithms.

The emphasis of the book is on a variety of medical applications such as gene expression data, meta-analysis and population pharmacokinetics. These applications are discussed in detail using real data from the medical literature.

The book offers an R package which enables the reader to use the methods for his/her needs.